B

Analyzing categorical data
Synopsis:-categorical data are encountered in many area of RS in human geography, notably those
where questiniary surveys are an important means of obtaining information
:- it focuses on how to undertake statistical analysis of search data most efficiently
:-it describes on how to explore trends within set of categorical variable and non-parametric techniques
(e.g the chis -square test) to evaluate hypothesis.
:- it explain how generalized linear model can be used to investigate more complex relationship between
categorical variable and discuss the circumstances in which these methods are likely to be particular
benefit.
Introduction:-the two variable data (nominal and ordinal) are discrete (that mean can take only certain
value) and are types of categorical data by contrast ratio data are continues and have metric properties.
It is always possible to convert data from higher level of measurement to the lower one .that mean ratio
to ordinal to nominal but not reverse direction (nominal to ordinal to ratio).
* if you are undertaking a questionnaires survey as a part of you dissertation. You may will find that
much of the data generated is categorical in form-
Ex yes/no respondent
Asked thier opinion
_agree
-disagree
-strong agree
-strong disagree
-neutral
*categorical variable may be also encountered in awdiw variety of others sources /e.g health services
recard or historical document.
*statistical analysis of such data need not cause your problems but does require atitudes care.this
partly because there are many possible techniques and thier appropriateness ends on the manner
measurement of the question.
* the another objective in analysis of categorical data should the relationship between variables are
examined in as through a manner as possible. Evaluate association between pairs of variable because chi
square multi variable by the sequence.
E.g 1st use of expiatory analysis then close confirmatory methods and change 3rd factories.
E.g the relationship of social case and attitude towards development of proposal various according the
area in which respondent lives and use generalize linear model e.g log linear logit method.
It uses interaction between variable data and its combination of other factor .
As generally the categorical data is provide guidance on the statistical analysis .the intention is not to
discuss individual techniques in a great detail through reference to source of further information will be
given instead to focus will be on broader issue that arise in
1. Exploratory investigation
2. Hypothesis testing
3. The contraction of statistical model
1. Exploratory analysis
It is a method of data analysis on the descriptive summary and graphical display of variable /detail see .
turkey in 11997 march 1998 Erickson and neuanchuk 1992
They can use to examine trend in data and are also effective at detecting outliers .that mean unusual
value like nominal variable.
It cheack the outlier and simply reflect coding and typing mistake and identify such errors can prevent
much frustration and need to repeat analysis.
If the outliers do not appear to be a mistake it is worthwhile to know that they exist because some value
are sensitive.
In the exploratory good starting paint is a distribution of value within the nominal data you can do in the
software to compute the number of observation in each categorical variable and perhaps to also express
the count of percentage.
Example –percent questionaries’ in respondent
-graphic from(eg bar charts )rather than text
In generally how to implement these type of summary and graphical techniques using spss are provided
by ryman and cramer (200)or miller et.al(2000).the book by town end/2002/ cover similar material to
the excel mini tab software packages.
The nominal variable can be investigated by cross tabulating data to for contingency table.
Example 12.1, 12.2, 12.3, 12.4 the contingency of connection of chapter six.
The survey respondent to the three area nor which were asked
1. CASTESSEY
2. SOTH PARK
3. WOOD CACK
By different way that mean local recycling center , transforming count to % of row total , by their way one of the
variable you can cross tabulating is thought to influence the other then MARSH /1998/ suggest that it is better to
construct contingency table so that the % sum to 100 within the categories of the exploratory of variable.
1. The tabulation use software /min tab URSSP/ extend three or more variables relationship of specialy
valuable given the difficulties with multiple causation and can founding that frequently encouraged in
human geography.
In third variable (recycling activity) when the values of the % or respondent in each area or housing type
combination who use recycling facility but if the third variable was a numerical it calculate mean or
median. Appropriate exploratory methods for ordinal variable differ according whether the data are
weakly or strongly ordered.
 Weakly ordered variable or ranked categories invariable high proportion of tide ranks by these
case best example is nominal.
 Strongly ordered or example contain few ties then scatter plot are a useful means of exploring
association between variable the geographic display and standing mistakes from the general
trend. Example table 12.1 and chapter 11 the relationship 3 or more ordinal variable.
BASIC CONFIRMATORY TECHNIQUES
 Once a set of data has been screened for outliers
 It is possible relationship between variable have been explored
 In the testing of hypothesis to determine weather trend apparent in sample of data example a
set of questionnaires response are also widely exist in a wide population.
 In the confirmatory the chaique there are two result are followed
1 statistical significant result -the trend in the sample data is sufficiently strong that there is only
a small probability or convention under 5% it ,occurred by chance
2 statistical weak significant result - the sample size of the sample has considerable influence
outcome
That mine the sample size of the large variable effect has weaker any trend need be quality as
statistically significance. Therefore real danger you don’t do limited ability.
In the case of nominal or ordinal variable the appropriate techniques are
1 non parametric
2 parametric
1 non parametric is provided by SIEGAL and CASTELLAN in 1988,GIBBANS 1993 and CONOVER in 1999
In these test have lower statistical efficiency that mean the power to detect an effect than their
parametric equivalent. But they also make fewer assumption about the parameters of population from
which sample are drown
it is more robust and consequently there can be situation. Example where frequency distribution are
heavily skewed.

B

Uploaded by

Copyright:

Available Formats

B

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

B

Uploaded by

Copyright:

Available Formats

Analyzing categorical data

:- it focuses on how to undertake statistical analysis of search data most efficiently

Asked thier opinion

3. The contraction of statistical model

Example –percent questionaries’ in respondent

-graphic from(eg bar charts )rather than text

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.